Deontic Policies for Runtime Governance of Agentic AI Systems TRAP: Benchmark for Task-completion and Resistance to Active Privacy-extraction Neural Additive and Basis Models with Feature Selection and Interactions When, Where, and How: Adaptive Binning for Tabular Self-Supervised Learning Reinforcement Learning Foundation Models Should Already Be A Thing